CoreWeave brings Vera Rubin NVL72 to cloud with 4.8x token generation speedup

NVIDIA and CoreWeave announced general availability of Vera Rubin NVL72 and Vera CPU on CoreWeave Cloud. Early inference tests show up to 4.8x higher total token throughput versus GB200 NVL72, while Vera CPU delivers 3x faster agentic sandbox startup.

• Vera Rubin NVL72 delivers up to 4.8x higher token throughput vs GB200 NVL72
• Vera CPU speeds up agentic sandbox startup by 3x
• A single rack holds 128 Vera CPUs (11,264 cores) for 11,000+ environments
• Cognition was first to deploy and scaled to thousands of GPUs in nine months